1
00:00:02,360 --> 00:00:10,040
Speaker 1: Bloomberg Audio Studios, Podcasts, radio News. I'm Stephen, Carol andvis

2
00:00:10,119 --> 00:00:12,440
Speaker 1: is Here's Why, where we take one news story and

3
00:00:12,560 --> 00:00:15,040
Speaker 1: explain it in just a few minutes with our experts

4
00:00:15,080 --> 00:00:16,040
Speaker 1: here at Bloomberg.

5
00:00:19,320 --> 00:00:22,520
Speaker 2: It's ten thirty pm in this AI party. It started

6
00:00:22,640 --> 00:00:25,680
Speaker 2: nine pm and that party goes to four am. And

7
00:00:25,760 --> 00:00:27,840
Speaker 2: the reality is like, look, this is going to be

8
00:00:27,880 --> 00:00:30,680
Speaker 2: a two to three year left in this bull cycle

9
00:00:30,760 --> 00:00:34,240
Speaker 2: for tech. The tech sector is very strong because artificial

10
00:00:34,240 --> 00:00:37,680
Speaker 2: intelligence is really a qualitative leap in the kind of

11
00:00:37,720 --> 00:00:41,199
Speaker 2: technology that we've had over the last several decades. You're

12
00:00:41,200 --> 00:00:45,160
Speaker 2: seeing an exponential growth of adoption and use of AI.

13
00:00:45,520 --> 00:00:47,920
Speaker 3: The number of applications that are going to be using

14
00:00:47,920 --> 00:00:49,600
Speaker 3: these AI is also growing.

15
00:00:50,120 --> 00:00:53,239
Speaker 1: Everyone has an opinion on where the AI frenzy is

16
00:00:53,280 --> 00:00:57,480
Speaker 1: going next. While optimism is rampant about the technology's potential,

17
00:00:57,840 --> 00:01:01,360
Speaker 1: more questions are now being asked about a running costs.

18
00:01:01,920 --> 00:01:06,640
Speaker 3: We are putting mostly chips silicon into these data centers

19
00:01:06,640 --> 00:01:10,360
Speaker 3: that have a lifespan of perhaps four years, So those

20
00:01:10,440 --> 00:01:12,560
Speaker 3: chips they appreciate very quickly.

21
00:01:12,720 --> 00:01:14,400
Speaker 1: Even in video, there's a new chip.

22
00:01:14,360 --> 00:01:17,080
Speaker 3: Every eighty months and it's ten times as powerful as

23
00:01:17,120 --> 00:01:19,040
Speaker 3: the earlier ones. The thing with the rally this year

24
00:01:19,160 --> 00:01:22,039
Speaker 3: is that almost every investor knows it's all going to

25
00:01:22,040 --> 00:01:25,200
Speaker 3: turn into pumpkins and mice at midnight. Only, as Buffett

26
00:01:25,240 --> 00:01:27,160
Speaker 3: would say, no one in the room as a clock.

27
00:01:27,680 --> 00:01:31,959
Speaker 1: Even with bumper results and bullish revenue forecasts, here's why

28
00:01:32,040 --> 00:01:39,040
Speaker 1: AI costs still worry investors. Tom McKenzie, who hosts Bloomberg

29
00:01:39,040 --> 00:01:42,640
Speaker 1: Tech You're Pumplerberg Television, joins me now for more. Tom.

30
00:01:42,959 --> 00:01:45,759
Speaker 1: The investor Michael Burry of Big Short fame is among

31
00:01:45,800 --> 00:01:48,920
Speaker 1: those who's worried about these future costs of AI and

32
00:01:49,080 --> 00:01:52,160
Speaker 1: data centers in particular. What's the concern?

33
00:01:52,800 --> 00:01:56,640
Speaker 3: Yeah, absolutely, Michael Burry putting on famously short positions, so

34
00:01:56,720 --> 00:01:59,840
Speaker 3: shortening the stocks of Nvidia and pallanteed before he wrapped

35
00:01:59,880 --> 00:02:05,320
Speaker 3: up his fund. His concern does focus on the depreciation

36
00:02:05,680 --> 00:02:08,920
Speaker 3: of some of these assets by assets. I'm talking about

37
00:02:08,960 --> 00:02:13,560
Speaker 3: specifically these AI chips, very expensive AI accelerators. Ninety percent

38
00:02:13,960 --> 00:02:17,760
Speaker 3: of the market share is dominated by Nvidio, so across

39
00:02:18,200 --> 00:02:21,440
Speaker 3: the sale of these chips and video has that significant

40
00:02:21,440 --> 00:02:24,880
Speaker 3: market gain versus its rivals. And the concern is that

41
00:02:25,080 --> 00:02:27,440
Speaker 3: as you get newer versions of these chips, the older

42
00:02:27,440 --> 00:02:32,000
Speaker 3: ones essentially become less valuable, and Michael Bari making the

43
00:02:32,080 --> 00:02:36,240
Speaker 3: argument that companies the hyperscalers, so the Microsofts and alphabets

44
00:02:36,400 --> 00:02:39,680
Speaker 3: and metas of the world, are not properly accounting for

45
00:02:39,800 --> 00:02:44,200
Speaker 3: how quickly these these assets depreciate. The other part of

46
00:02:44,200 --> 00:02:46,720
Speaker 3: the concern and kind of ties into this that you

47
00:02:46,800 --> 00:02:49,800
Speaker 3: hear voice from the skeptics around the AI bubble is

48
00:02:49,800 --> 00:02:53,080
Speaker 3: that there are comparisons, they say, with what happened in

49
00:02:53,120 --> 00:02:55,720
Speaker 3: the late nineteen nineties, nineteen ninety nine early two thousand,

50
00:02:55,720 --> 00:02:58,880
Speaker 3: the dot com bubble, when it was the telecom equipment

51
00:02:58,919 --> 00:03:04,040
Speaker 3: makers that leading up to all of the online expectations

52
00:03:04,080 --> 00:03:06,520
Speaker 3: around how our digital economy was going to change, spent

53
00:03:06,720 --> 00:03:10,080
Speaker 3: huge amounts of money on building the infrastructure to power

54
00:03:10,720 --> 00:03:14,919
Speaker 3: the dot com era and ended up losing a lot

55
00:03:14,960 --> 00:03:17,880
Speaker 3: of money because the gains didn't come as quickly, the

56
00:03:17,919 --> 00:03:21,399
Speaker 3: technology didn't evolve as rapidly as they had expected. Of course,

57
00:03:21,440 --> 00:03:23,720
Speaker 3: on the back of that you did get some very

58
00:03:23,760 --> 00:03:27,399
Speaker 3: significant players like Amazon who came through the dot com

59
00:03:27,440 --> 00:03:29,960
Speaker 3: bubble and of course now remain one of the most

60
00:03:30,040 --> 00:03:32,519
Speaker 3: valuable companies on the planet. But there was a lot

61
00:03:32,560 --> 00:03:34,400
Speaker 3: of capital, there was a lot of investment that was

62
00:03:34,480 --> 00:03:37,200
Speaker 3: burnt in that process. And so that is another comparison

63
00:03:37,200 --> 00:03:39,320
Speaker 3: that people are making. It's the depreciation around the assets

64
00:03:39,360 --> 00:03:42,040
Speaker 3: and the chips that they're worried about, but also comparisons

65
00:03:42,080 --> 00:03:44,640
Speaker 3: with what happened during the dot com era and the

66
00:03:44,680 --> 00:03:47,000
Speaker 3: pain that was felt by those telecom equipment makers that

67
00:03:47,280 --> 00:03:50,800
Speaker 3: sunk so much money into which they accumulated huge losses.

68
00:03:51,400 --> 00:03:55,000
Speaker 1: So how are the big aiplayers thinking about these casts

69
00:03:55,120 --> 00:03:55,720
Speaker 1: at the moment.

70
00:03:56,400 --> 00:04:00,760
Speaker 3: So push back to the depreciation argument would come from

71
00:04:00,840 --> 00:04:03,080
Speaker 3: in video and we've heard that recently from the CEO

72
00:04:03,160 --> 00:04:06,520
Speaker 3: Jensen Huang, and he's made the case that in fact,

73
00:04:06,960 --> 00:04:10,360
Speaker 3: even their older AI chips, one of their older versions

74
00:04:10,400 --> 00:04:14,000
Speaker 3: is called Hopper, has a lifespan of about six years

75
00:04:14,240 --> 00:04:16,960
Speaker 3: and is very versatile. So you can use it not

76
00:04:17,200 --> 00:04:20,520
Speaker 3: just for the training of these large language models, but

77
00:04:20,560 --> 00:04:23,160
Speaker 3: for the post training and for the inference that's when

78
00:04:23,160 --> 00:04:26,560
Speaker 3: they're actually being used by us, by consumers and by enterprise,

79
00:04:27,040 --> 00:04:29,240
Speaker 3: and so you can move them around. They have different

80
00:04:29,240 --> 00:04:32,520
Speaker 3: functions and therefore they actually have a longer life span

81
00:04:32,960 --> 00:04:35,760
Speaker 3: than some of the skeptics are suggesting. And our own

82
00:04:35,800 --> 00:04:38,560
Speaker 3: analysis suggests that those Hopper chips, those older varieties of

83
00:04:38,640 --> 00:04:40,440
Speaker 3: chips have a life span of about six years and

84
00:04:40,520 --> 00:04:44,400
Speaker 3: are fully utilized by most of the companies that own those,

85
00:04:44,400 --> 00:04:49,320
Speaker 3: So that does address some of that concern. The question

86
00:04:49,720 --> 00:04:53,039
Speaker 3: going forward to what extent these companies are going to

87
00:04:53,040 --> 00:04:57,400
Speaker 3: be able to find products that match the investments that

88
00:04:57,440 --> 00:05:01,280
Speaker 3: they are syncing into the AI infrastructure story. A Bane

89
00:05:01,279 --> 00:05:04,280
Speaker 3: Capital came out with a report recently suggesting that by

90
00:05:04,320 --> 00:05:08,920
Speaker 3: twenty thirty, the hyperscalers and other AI giants would have

91
00:05:09,000 --> 00:05:12,040
Speaker 3: to be turning around revenues of about two trillion dollars

92
00:05:12,640 --> 00:05:15,320
Speaker 3: and that right now there's a huge gap hundreds of

93
00:05:15,320 --> 00:05:17,640
Speaker 3: billions of dollars in terms of the gap between the

94
00:05:17,760 --> 00:05:22,400
Speaker 3: investments into the AI infrastructure and the actual revenues that

95
00:05:22,440 --> 00:05:26,120
Speaker 3: are coming about as customers and as enterprises and companies

96
00:05:26,560 --> 00:05:28,839
Speaker 3: use the end product. So the go to market, the

97
00:05:28,839 --> 00:05:31,000
Speaker 3: product fit is going to be really, really important. And

98
00:05:31,040 --> 00:05:33,440
Speaker 3: what the big AI players say, whether or that is

99
00:05:33,480 --> 00:05:36,440
Speaker 3: the Hyperscalers again, the likes of Meta and in Alphabet

100
00:05:36,440 --> 00:05:38,560
Speaker 3: and Amazon say, all the likes of open A and

101
00:05:38,560 --> 00:05:40,279
Speaker 3: anthrop because we're going to be in this world of

102
00:05:40,400 --> 00:05:44,160
Speaker 3: agentic AI. We're going to have AI agents booking our holidays,

103
00:05:44,520 --> 00:05:48,040
Speaker 3: checking up on our healthcare, finding good schools and universities

104
00:05:48,080 --> 00:05:50,040
Speaker 3: for our students. All those kind of things are going

105
00:05:50,080 --> 00:05:53,000
Speaker 3: to come together. Enterprises are going to be embedding AI

106
00:05:53,560 --> 00:05:55,599
Speaker 3: much more than they already are. We're only in the

107
00:05:55,640 --> 00:05:58,120
Speaker 3: first opening stages of that would be the argument. And

108
00:05:58,160 --> 00:06:01,359
Speaker 3: then there's the sovereign AI story where different countries, and

109
00:06:01,400 --> 00:06:02,960
Speaker 3: we're seeing that in the Middle East, but also in

110
00:06:02,960 --> 00:06:05,799
Speaker 3: Europe as well, and Japan are investing heavily to ensure

111
00:06:05,800 --> 00:06:08,599
Speaker 3: that they have their own AI infrastructure, AI clouds that

112
00:06:08,640 --> 00:06:11,159
Speaker 3: will very early in that story as well. Those are

113
00:06:11,200 --> 00:06:14,360
Speaker 3: all the cases that the big AI players would underscore

114
00:06:14,560 --> 00:06:17,680
Speaker 3: in terms of why this is going to be driving

115
00:06:17,760 --> 00:06:21,160
Speaker 3: momentum going forward, at least through twenty twenty six. Our

116
00:06:21,200 --> 00:06:23,880
Speaker 3: own team at Bloomberg Intelligence say the end of twenty

117
00:06:23,880 --> 00:06:25,520
Speaker 3: twenty six is going to be a question mark as

118
00:06:25,520 --> 00:06:28,520
Speaker 3: to whether or not investors continue to have patients. Will

119
00:06:28,520 --> 00:06:31,880
Speaker 3: they continue to invest in the hyperscalers if they're not

120
00:06:31,920 --> 00:06:34,919
Speaker 3: saying real material terms, if that product fit and that

121
00:06:35,040 --> 00:06:38,520
Speaker 3: custom use isn't there in a really really significant way.

122
00:06:38,600 --> 00:06:41,200
Speaker 3: So I think the patients of investors and to what

123
00:06:41,279 --> 00:06:44,920
Speaker 3: extent they can continue to lean into the hyperscalers as

124
00:06:44,960 --> 00:06:47,120
Speaker 3: they spend these huge amounts is going to be a

125
00:06:47,160 --> 00:06:48,880
Speaker 3: key question mark, and our own team think that that's

126
00:06:48,920 --> 00:06:50,520
Speaker 3: really going to come to the fore at the end

127
00:06:50,520 --> 00:06:53,000
Speaker 3: of twenty twenty six, they'll need to answer that question.

128
00:06:53,200 --> 00:06:55,679
Speaker 3: They've spent the hyper scalers have spent about three hundred

129
00:06:55,720 --> 00:06:59,320
Speaker 3: billion dollars on air infrastructure this year, and the projection

130
00:06:59,440 --> 00:07:01,720
Speaker 3: is that they could be according to Vidia in the video,

131
00:07:01,760 --> 00:07:04,840
Speaker 3: sees the hyperscalar spinning upwards of about six hundred billion

132
00:07:04,880 --> 00:07:05,680
Speaker 3: dollars next year.

133
00:07:06,120 --> 00:07:07,560
Speaker 1: One of the things that occurs to me in this

134
00:07:07,640 --> 00:07:10,120
Speaker 1: as well, as we're talking about some of the world's

135
00:07:10,280 --> 00:07:14,600
Speaker 1: most valuable companies, they have massive cash piles in a

136
00:07:14,600 --> 00:07:18,000
Speaker 1: lot of cases, Why is there concern at all about

137
00:07:18,040 --> 00:07:20,440
Speaker 1: how they're going to pay for this given their revenue

138
00:07:20,440 --> 00:07:21,880
Speaker 1: streams and how much money they have.

139
00:07:22,280 --> 00:07:25,120
Speaker 3: You're absolutely right. So when we talk about the hyperscalers,

140
00:07:25,560 --> 00:07:30,200
Speaker 3: these are companies with massive balance sheets and huge cash reserves.

141
00:07:30,240 --> 00:07:34,160
Speaker 3: These are incredibly profitable businesses that come through with very

142
00:07:34,200 --> 00:07:39,120
Speaker 3: strong earnings. These are not nonprofitable major punts and risky

143
00:07:39,160 --> 00:07:41,280
Speaker 3: parts of the market. These are not companies that no

144
00:07:41,280 --> 00:07:44,320
Speaker 3: one's heard of. They're making real product they're selling it

145
00:07:44,360 --> 00:07:48,440
Speaker 3: to customers, and they've been doing that for decades. Microsoft, Alphabet, Meta,

146
00:07:48,480 --> 00:07:51,080
Speaker 3: and Amazon. They have that balance sheet strength, they have

147
00:07:51,200 --> 00:07:55,920
Speaker 3: that cash on hand. The concern then is around other

148
00:07:55,960 --> 00:07:57,920
Speaker 3: parts of this ecosystem. So if you can think about

149
00:07:57,960 --> 00:08:01,000
Speaker 3: it in different baskets, you have those big ticket blue

150
00:08:01,080 --> 00:08:04,480
Speaker 3: chip names in one basket, and then you have maybe

151
00:08:04,520 --> 00:08:07,280
Speaker 3: neo clouds in the other basket. These are the core

152
00:08:07,320 --> 00:08:11,560
Speaker 3: weaves or the n clouds companies that lease out data

153
00:08:11,600 --> 00:08:13,880
Speaker 3: centers to some of these hyper scalers, and some of

154
00:08:13,880 --> 00:08:16,840
Speaker 3: the large language models who have business models that are

155
00:08:16,960 --> 00:08:20,720
Speaker 3: less proven than the hyperscalers. Then another bucket would be

156
00:08:21,560 --> 00:08:23,960
Speaker 3: maybe some of the key large language models themselves, the

157
00:08:23,960 --> 00:08:26,520
Speaker 3: Opening Eyes and the Anthropics, that are losing money on

158
00:08:26,560 --> 00:08:28,960
Speaker 3: an annual basis, even as they're seeing a lot of

159
00:08:28,960 --> 00:08:34,240
Speaker 3: growth and revenues increase year on year, they're still not profitable.

160
00:08:34,480 --> 00:08:36,600
Speaker 3: So you can break it down into different categories in

161
00:08:36,640 --> 00:08:39,840
Speaker 3: terms of the level of risk. But even amongst the

162
00:08:39,840 --> 00:08:42,520
Speaker 3: big publicly listed companies with those strong balance sheets, you

163
00:08:42,600 --> 00:08:45,480
Speaker 3: have seen examples of then tapping the public markets and

164
00:08:45,600 --> 00:08:47,800
Speaker 3: raising debt on the public markets, and so far that's

165
00:08:47,840 --> 00:08:50,960
Speaker 3: been well received. By the markets. But how long is

166
00:08:51,000 --> 00:08:53,080
Speaker 3: that going to continue? And to what extent is the

167
00:08:53,200 --> 00:08:56,760
Speaker 3: leverage that now these companies are starting to tap into

168
00:08:57,280 --> 00:09:00,240
Speaker 3: going to be acceptable to investors? And again, I think

169
00:09:00,280 --> 00:09:02,520
Speaker 3: you have to put a different framework over the different

170
00:09:02,559 --> 00:09:05,600
Speaker 3: companies in terms of how you answer that question. Then

171
00:09:05,640 --> 00:09:08,560
Speaker 3: there's the circularity of the financing, so open Ai, for example,

172
00:09:08,600 --> 00:09:11,840
Speaker 3: doing deals with Nvidia, and Nvidia investing in open Ai,

173
00:09:12,000 --> 00:09:14,960
Speaker 3: and in response to that, open Ai committing to buying

174
00:09:14,960 --> 00:09:18,640
Speaker 3: a certain number of chips from Nvidia. Those circular financing deals,

175
00:09:18,720 --> 00:09:21,200
Speaker 3: as they've been described by some have also caused some

176
00:09:21,280 --> 00:09:25,520
Speaker 3: concern as all of these companies becoming increasingly enmeshed and

177
00:09:25,640 --> 00:09:29,040
Speaker 3: intertwined in terms of their deals and their investments on

178
00:09:29,080 --> 00:09:30,960
Speaker 3: what is a bet on the future and how the

179
00:09:30,960 --> 00:09:31,640
Speaker 3: future evolves?

180
00:09:32,000 --> 00:09:34,280
Speaker 1: Had an expensive one of that, Tom McKenzie, thank you

181
00:09:34,400 --> 00:09:37,240
Speaker 1: very much for joining us, host of Bloomberg Tech Europe

182
00:09:37,480 --> 00:09:41,200
Speaker 1: on Bloomberg Television. For more explanations like this from our

183
00:09:41,240 --> 00:09:44,160
Speaker 1: team of three thousand journalists and analysts around the world,

184
00:09:44,240 --> 00:09:47,559
Speaker 1: go to Bloomberg dot com slash Explainers. I'm Stephen Carroll.

185
00:09:47,720 --> 00:09:50,160
Speaker 1: This is Here's why. I'll be back next week with more.

186
00:09:50,360 --> 00:09:51,120
Speaker 1: Thanks for listening,